Wessen randi (cd)
-
Upload
nasapmc -
Category
Technology
-
view
14.077 -
download
0
Transcript of Wessen randi (cd)
1
Market-Based Systems:Bidding your way to the
Launch Pad
Dr. Randii R. WessenJet Propulsion Laboratory
California Institute of TechnologyDr. Dave Porter
Interdisciplinary Center for Economic ScienceGeorge Mason University
National Aeronautics and Space AdministrationJet Propulsion LaboratoryCalifornia Institute of Technology
2
Agenda
• Introduction• Cassini Science Instrument Development• Space Shuttle Manifests• LightSAR Mission Planning• Market-Based Systems Applied to NASA
Resource Allocation Problems• Conclusions
3
Introduction
Problem - How to allocate scarce resources among many users?♦ Users do not have an incentive to reveal accurate information.
Current Approach:♦ Benevolent Dictator - Impartial individual making decisions for
user community. ♦ Committee - “Consensus” reached by user representatives.
New Approach:♦ Market-based systems - Decentralized incentive-based
decision-making at the user level.
4
Market-based systems use “rights” and “trades” to resolve conflicts.♦ Users:
– Own clearly defined resources.– Decide the importance of their resources.– Enhance their own position by exchanging resources
among themselves.
Introduction (cont.)
5
Market-based systems require that an “economy” be created.To create an economy, the following must be defined:♦ Resources to be allocated.♦ Resource ownership rights.♦ Rules for making and keeping track of trades.
Introduction (cont.)
6
Market-based systems have two major phases:♦ Initial Allocation (Primary Market) - Users use “currency” to
establish property rights over resources.♦ Aftermarket (Secondary Market) - Users can voluntarily
trade their resources with others to enhance their value.
Introduction (cont.)
7
Examples of Market-based systems used in private industry:
♦ Best Buy♦ Corning♦ Federal
Communications Commission
♦ General Electric♦ Google♦ Hewlett-Packard♦ Koch Industries♦ Microsoft
♦ RECLAIM Program in Southern California
♦ Sandia National Laboratories
♦ Sears Logistic Services♦ University of Chicago
Business School♦ University of Michigan♦ University of San Diego
Introduction (cont.)
8
The Cassini Instrument Development
9
Cassini Instrument Development
Typical instrument development for planetary missions require:♦ Project’s to submit a Request for Proposals. ♦ Project’s acceptance of Principal Investigators’ proposals.
Data from previous missions show that instrument mass and cost growths are:♦ Almost always positive. ♦ Can be very large.
10
Cassini Instrument Dev. (cont.)
Percent Growth for Past Space Missions
11
In 1992 a Market-based Aftermarket, known as the Cassini Resource Exchange (CRE), was developed.♦ Based on experiments conducted at Caltech with student
subjects and science personnel.
Instrument mass, power, data rate, and funding were allowed to be traded.♦ The CRE opened in 1993 and closed in 1995.
– 29 successful trades were made, all but two involved money and mass.
Cassini Instrument Dev. (cont.)
12
0
2
4
6
8
10
12
14
0 20 40 60 80 100 120 140 160 180 200
Week Number Starting in FY93
% R
ate
Cassini Trade
1 Year T-Bill
Example of a money-market trade:$200k this FY in return for $212k next FY = 6%.
16 contracts, worth $4M, were traded at anaverage rate of 8.5%.
Fig 2. Money Market Activity
Cassini Instrument Dev. (cont.)
13
Mass-market trades:♦ Mass started at $105k/kg and fell to $5k/kg.♦ 11 contracts for a total of 12 kg were traded.
Figure 5: Mass Market Activity
0
20
40
60
80
100
120
0 50 100 150 200
Week Number Starting in FY93
$k/K
g
Cassini Instrument Dev. (cont.)
14
Cost for the entire science payload grew by <1%.
Figure 6: Cassini's Instrument Cost Growth
0
10
20
30
40
50
60
CAPS
CDA
CIRS
INMS ISS MA
G
MIMI
RADA
R
RPW
S
RSS
UVIS
VIMS
Investigations
Tota
l Cos
t, $
M
LOA Cost
Final Cost
Cassini Instrument Dev. (cont.)
15
Cassini Instrument Dev. (cont.)
Mass for the entire science payload decreased by 7%.
Figure 7: Cassini's Instrument Mass Growth
0
10
20
30
40
50
60
70
CAPS
CDA
CIRS
INMS ISS MA
G
MIMI
RADA
R
RPW
S
RSS
UVIS
VIMS
Investigations
Mas
s, k
g. LOA Mass
Final Mass
16
Cassini Instrument Dev. (cont.)
In 1998 a Market-based system was developed for Cassini science planning.♦ Web-based application.
– Password protected.♦ Used to allocate observation time, CDS words and data volume.♦ Compatible with JPL’s Mission Sequence Software System
(e.g., APGEN)
17
Cassini Instrument Dev. (cont.)
Cassini science planning bid sheet.
18
Cassini Instrument Dev. (cont.)
• Cassini Science Planning Tool.
♦ Timeline is always conflict-free.
♦ Provides user feedback to get observation requests accepted.
♦ This system was never tested.
19
Space Shuttle Manifests of
Secondary Payloads
20
Space Shuttle Manifests
Typical approach for manifesting Space Shuttles Secondary Payloads require:♦ NASA User Codes to submit requests for payload lockers.♦ Manifestors at NASA Headquarters allocate lockers in such
a way to:– Utilize Space Shuttle resources to capacity.– Be “fair” to the User Codes.
21
Space Shuttle Manifests (cont.)
Multiple meetings required to:♦ Present current manifest.♦ Explain why some payloads were included while others
were not.♦ Obtain payload trade-off information.
Manifestors required to re-manifest secondary payloads after each meeting.
22
Space Shuttle Manifests (cont.)
In the summer of 1996 a Market-based Initial Allocation experiment was developed for NASA’s Office of Space Utilization.♦ Market-based approach was compared to a simple ranking
approach.♦ Secondary payload lockers, energy, and crew hour
requirements were allowed to be requested.
23
Space Shuttle Manifests (cont.)
Office of Space Utilization personnel performed the simple ranking manifest.♦ Results were compared to market-based results.
Real payload data was used.♦ A “science return” value and payload rank were included to
measure the caliber of the manifest.
USER P/L LOCKERS WATT-HRS CREW-HRS SCIENCE RETURN
Code U MGBX-01 6 237 56 100Code U CGBA-04 4 136 5.166 70Code U CPCG-07 1 128 5.833 55Code U CPCG-08 1 128 2 45Code U PCG-TES-02 3 115 0 40
12233
RANK
24
Space Shuttle Manifests (cont.)
75
80
85
90
95
100
% o
f C
apac
ity
Use
d
Lockers Watt-Hrs Crew-Hrs
Resource UtilizationRankingPoints
Comparison of resource utilization between asimple ranking vs a market-based approach.
ManualMBS
25
Space Shuttle Manifests (cont.)
-30
-20
-10
0
10
20
30
40
% E
xces
s R
etu
rn
Code U Code S Code X Code F DoD
Additional Science Return Relative to Baseline
Year 1 Year 2Total
Point carry-over 0 37 30 3 1Year 1 to Year 2
Comparison of science value for a simple rankingvs a market-based approach.
26
Space Shuttle Manifests (cont.)
Market-based systems have already handled the following type of problem:♦ Manifest space shuttles middeck lockers efficiently ♦ 6 Users♦ 47 variables/payload♦ 1,500,000 payload requests♦ Solutions found in an average of 2.5 minutes
27
Space Shuttle Manifests (cont.)
Implementation of a market-based system is “on-hold” until:♦ The full impact of the International Space Station’s
requirements on middeck lockers can be determined.
♦ “Nominal” Space Shuttle missions resume.
28
LightSARMission Planning
29
LightSAR Mission Planning
LightSAR was a NASA initiative to develop a low-cost Earth-imaging RADAR satellite.Typical approach for mission planning:♦ Submit data take requests.♦ Create integrated timeline.♦ Multiple meetings to resolve conflicts.♦ Re-Integrate timeline.
30
LightSAR Mission Planning (cont.)
In the winter of 1997 a Market-based Initial Allocation experiment was developed forLightSAR mission planning.♦ Users were assigned a specific data acquisition type (e.g.
Hi-Res Strip, Spotlight, Interferometry, Dual Polarimetry, etc.).
31
LightSAR Mission Planning (cont.)
A data request, rank and “science return”value were included with each request to measure the caliber of the manifest.
Dual Polarimetry Requests
Location Orbit
Number
Data Take
Number
Rank Value
Vietnam 1 1 1 60
Kuala Lumpar 1 3 2 45
Indonesia 1 4 2 35
Cambodia 2 3 3 10
32
LightSAR Mission Planning (cont.)
In a market-based approach, Users bid for a “higher priority” request.♦ The higher the priority, the greater the probability of
getting into the timeline.♦ A Vickrey Auction was used to provide “incentives” for the
Users to be forthright about their bids.– In a Vickrey Auction “winners”pay the runner-up price.
33
LightSAR Mission Planning (cont.)
Comparison of the science value between asimple ranking, a simple market, and a market using a Vickrey Auction.
99
100
101
102
103
104
105
106
107
108
% o
f D
raft
Ou
tco
me
Subject 1 Subject 2 Subject 3 Subject 4 Cumulative
Science Values
Simple MarketPriority MarkeSimple Market Market w/ Vickrey
Science value using a simple ranking approach
34
LightSAR Mission Planning (cont.)
35
LightSAR Mission Planning (cont.)
36
LightSAR Mission Planning (cont.)
37
LightSAR Mission Planning (cont.)
38
Experiments revealed only minor operational problems.♦ Lack of experience seemed to be the largest.
LightSAR was canceled.A new RADAR mission called INSAR has begun concept studies.♦ The work on LightSAR can be used for INSAR and will be
pursued.
LightSAR Mission Planning (cont.)
39
Market-Based Systems Applied to NASA Resource
Allocation Problems
40
NASA Market-Based Systems
Never performed experiments with users.
2005 April2005 JanDeep Space NetworkAntenna Allocation
Never performed experiments with users.
2002 Jan2000 JunAlternate Workforce Allocations
Never performed experiments with users.
On Hold2000 SeptInternational Space Station Manifests
Mission canceled.1999 Dec1999 AprilMars’01 Science Planning
Never performed experiments with users.
1999 Mar1998 AprilCassini Science Planning
Mission canceled.1998 Sept1997 DecLightSAR Mission Planning
Never made it past experiments.1997 Sept1996 DecSpace Shuttle Manifests
Successfully controlled mass and cost growth.
1995 April1993 SeptCassini Instrument Development
StatusEndStartTask
41
Conclusions
42
Cassini, Space Shuttle, and LightSAR benefited (or could benefit) from the use of a market-based process.Resource allocation problems can be solved with the use of market-based systems.♦ Data shows that market-based systems can produce results:
– As good (or better) than the current processes.– Are reached quicker and with a smaller workforce.
Conclusions
43
The strengths of market-based systems are that they:♦ Move the decision making process back to the individuals
that have the information, namely the users.♦ Are electronically based on the Web and thus can be
globally distributed.♦ Remove the need for multiple meetings and appeals.
Conclusions
44
One perceived short coming of market-based systems is their initial allocation.♦ However, all processes have initial allocation problems.♦ Better to solve one large problem early then multiple smaller
problems late in a given process.
Most individuals outside of experimental economics do not appreciate the power ofmarket-based systems.
Conclusions
45
In order to successfully implement a market-based system it is recommended that it be used in parallel with the current approach to “gently” help users understand its application.There will be resistance to change to a new system, even a more efficient one.♦ Winners are not sure that they’ll do as well as compared to
the current process.♦ Losers are not sure that they’ll do better as compared to the
current process.♦ Integrators lose their jobs.♦ Managers will express a loss of control.
Conclusions
46
Top 5 “bad” reasons for not using a market-based system:
5. “Market-based systems can solve resource allocation problems, but my problem is different.”
4. “We don’t have a resource allocation problem, we just just have a scheduling problem.”
3. “There’s no way to come up with an initial allocation of resources.”
2. “Too many people will lose their jobs.”1. “We don’t need a market-based system to solve our problems.
We’ll just solve it in a collegial manner.”
Conclusions